Analytical Biochemistry 283, 207213 (2000) doi:10.1006/abio.2000.4641, available online at http://www.idealibrary.

com on

A New Method for the Determination of Stability Parameters of Proteins from Their Heat-Induced Denaturation Curves
Sushma Yadav and Faizan Ahmad 1
Department of Biosciences, Jamia Millia Islamia, Jamia Nagar, New Delhi 110 025, India

Received January 13, 2000

A new method has been developed for determining the stability parameters of proteins from their heatinduced transition curves followed by observation of changes in the far-UV circular dichroism (CD). This method of analysis of the thermal denaturation curve of a protein gave values of stability parameters that not only are identical to those measured by the differential scanning calorimetry (DSC), but also are measured with the same error as that observed with a calorimeter. This conclusion has been reached from our studies of the reversible heat-induced denaturation of lysozyme and ribonuclease A at various pH values. For each protein, the conventional method of analysis of the conformational transition curve, which assumes a linear temperature dependence of the preand posttransition baselines, gave the estimate of H mvan (enthalpy change on denaturation at T m, the midpoint of denaturation) which is signicantly lower cal , the value obtained from DSC measurethan H m ments. However, if the analysis of the same denaturation curve assumes that a parabolic function describes the temperature dependence of the pre- and posttransition baselines, there exists an excellent van cal and H m of the protein. The agreement between H m latter analysis is supported by the far-UV CD measurements of the oxidized ribonuclease A as a function of temperature, for the temperature dependence of this optical property of the protein is indeed nonlinear. Furthermore, it has been observed that, for each protein, the constant-pressure heat capacity change (C p) van versus T m is indedetermined from the plots of H m pendent of the method of analysis of the transition curve. 2000 Academic Press

Key Words: protein stability; vant Hoff analysis; ribonuclease A; lysozyme; thermal denaturation.

One of the estimates of protein stability is G D , the Gibbs energy change associated with the process, N (native) conformation 7 D (denatured) conformation under physiological conditions usually taken as neutral dilute buffer (or water) at 25C. A knowledge of an accurate estimate of G D is essential, for nearly all theoretical and experimental aspects of protein folding relate in some way to this quantity. G D is mostly determined from the measurements of the reversible heat-induced denaturation of proteins using calorimetric and equilibrium methods (1). The differential scanning calorimetry (DSC) 2 provides direct estimates of the constant-pressure heat capacity change ( C P) and H m, the enthalpy change at T m, the midpoint of denaturation. These three thermodynamic parameters are then used to estimate G D using the Gibbs-Helmoltz equation (see below). On the other hand, the equilibrium method involves (i) measurements of the conformational transition between N and D states followed by observation of changes in a suitable structural property of the protein, and (ii) analysis of the transition curve for the equilibrium constant ( K D). Hence this method is called the equilibrium method. A knowledge of the mathematical function(s) that is (are) most appropriate to describe the temperature dependencies of the pre- and posttransition baselines is a prerequisite for the analysis of the conformational transition curve (see below). Furthermore, the equilibrium method is a noncalorimetric technique, and all
2 Abbreviations used: RNase-A, ribonuclease A; DSC, differential scanning calorimetry; CD, circular dichroism.

1 To whom correspondence should be addressed. Fax: 91-11-5791351. E-mail: Faizana@del3.vsnl.net.in.

0003-2697/00 $35.00 Copyright 2000 by Academic Press All rights of reproduction in any form reserved.

207

208

YADAV AND AHMAD

the derived thermodynamic parameters should be validated against calorimetric data. The conventional method of analysis of optical transition curves assumes that the temperature dependencies of the pre- and posttransition baselines of a protein are linear. Recently, we have shown that (i) use of this assumption in the analysis of transition curves measured by the absorption properties of proteins gave a G D value which is signicantly lower than that obtained from DSC; (ii) the discrepancy between G D estimates from DSC and equilibrium methods is removed if a parabolic function describes the temperature dependencies of the absorption properties of the proteins in the pre- and posttransition regions; and (iii) the temperature dependencies of the absorption properties of model compounds are indeed nonlinear (2). A question arises: does this apply to another optical property of proteins, namely, far-UV circular dichroism (CD) which is an excellent technique for measuring heat-induced denaturation curves of proteins (3)? In order to understand this question, we have been conducting systematic studies of the heat-induced denaturations of several proteins. This is the rst such study in which we report thermodynamic stability parameters obtained from the denaturation of ribonuclease A (RNase-A) and hen egg lysozyme. It has been observed that (i) the conventional method of analysis, which assumes a linear temperature dependence of the pre- and posttransition baselines, gave a signicantly lower estimates of G D than that from DSC measurements; (ii) agreement between stability parameters is excellent if the pre- and posttransition baselines are described by a parabolic function; and (iii) the temperature dependence of the CD of denatured proteins is indeed nonlinear.
MATERIALS AND METHODS

containing 0.1 M KCl. For the various pH ranges, the buffers used were 0.05 M KCl-HCl buffer (pH 1.50 2.50), 0.05 M Gly-HCl buffer (pH 2.50 3.20), and 0.03 M cacodylic acid buffer (pH 4.50 6.00). Since the pH of the protein solution may change on heating, the pH of the solution was therefore measured after the denaturation experiment. It has been observed that the change in pH is not signicant. Thermal denaturation studies were carried out in a Jasco J-715 spectropolarimeter equipped with a peltier-type temperature controller (PTC-348 WI), with a heating rate of 1C/min. Change in CD at 222 nm of the protein solution (concentration in the range 0.3 0.5 mg/ml) was measured in the temperature range 20 85C. About 650 data points were collected. After denaturation, the sample was immediately cooled to measure reversibility of the reaction. All solution blanks showed negligible change in ellipticity with temperature and were, therefore, neglected during data analysis. The raw CD data were converted into the mean residue ellipticity [] (expressed in deg cm 2 dmol 1) using the relation,

M 0 , 10 lc

[1]

where is the observed ellipticity in millidegrees at wavelength , M 0 is the mean residue weight of the protein, c is the protein concentration (mg/cm 3), and l is the pathlength (in cm).
RESULTS AND DISCUSSION

Bovine pancreatic RNase-A (type XII-A), oxidized RNase-A (type XII-AO), and hens egg lysozyme (grade I) were obtained from Sigma. Since all the proteins gave single band on SDS-PAGE, no further purication was done. Since sodium salt of cacodylic acid, glycine, and KCl, purchased from Aldrich Chemical Co., and reagents and solvents used for SDS-PAGE, purchased from SRL Chemical (India), were of analytical grade, they were used without further purication. Stock solutions of proteins were prepared by exhaustive dialysis against 0.1 M KCl (pH 7.0) at 4C. These solutions were ltered using 0.45-m Millipore lter paper, and their concentrations were determined from the absorbance of an appropriately diluted aliquots using molar absorption coefcient values of 39000 M 1 cm 1 at 280 nm for lysozyme (4) and 9800 M 1 cm 1 at 277.5 nm for RNase-A (5). All solutions for CD measurements were prepared in the appropriate buffer

Recently, we have estimated stability parameters of several proteins using absorption spectroscopy (2). We have shown that, for a protein, these estimates obtained from the conventional method of analysis of transition curves are signicantly smaller than those from DSC measurements. Several possible origins of differences between equilibrium and calorimetric estimates of stability parameters have been considered (2). A possibility which was not considered earlier (2) is the fact that the observed denaturation curve of the protein tells us only about the local changes in the environment of the chromophore; i.e., it does not reect conformational changes in the protein segments devoid of a chromophore, whereas DSC measurements reects changes in heat capacity of the whole protein molecule. Thus a valid comparison between equilibrium and calorimetric estimates of a thermodynamic parameter should involve the measurements of the denaturation curve followed by an observation of changes in a protein property which is sensitive to conformational changes in all parts of the protein molecule. Far-UV CD is indeed such a property, for the CD spectrum of each element of the native conformation, be it -helix,

THERMAL STABILITY OF PROTEINS

209

FIG. 1. Thermal denaturation curves of lysozyme (A) and RNase-A (B) at different pH values which are given in the gure. Assuming that the temperature dependence of [] 222 is linear (L) and parabolic (P), the pre- and posttransition baselines of lysozyme (A) were drawn using data at pH values 3.18 and 1.52, respectively, and those of RNase-A (B) were drawn using data at pH 5.95. The inset in (B) shows the [] 222 of the oxidized RNase-A. In order to maintain the clarity not all data points are shown on the curves.

-stucture, -turn, or unordered conformation, is different from that of the unfolded protein [see Fig. 1 in (3)]. We present here the method that may be used to obtain stability parameters from the analysis of heatinduced chiroptical transition curves of proteins. The heat-induced denaturation of two model proteins, namely, lysozyme and RNase-A was followed by observing changes in [] 222, which is considered to be an index of protein secondary structure (3, 6). Figure 1 shows the denaturation results of lysozyme (A) and RNase-A (B) at four different pH values. If the following conditions are satised, these optical data of each transition curve can be converted into equilibrium constants using the relation,
y T y N T , y D T y T

K D T

[2]

where y ( T ) is the experimentally observed far-UV CD property of the protein at T K, y N( T ) and y D( T ) are the CD properties of the native and denatured molecules at T K. These conditions are: (i) the denaturation is reversible; (ii) the denaturation is a two-state type; (iii) [ y ( T ) y N( T )] and [ y D( T ) y ( T )] represent the concentrations of denatured and native molecules, respectively; (iv) temperature dependencies of y N and y D

are accurately known; and (v) thermodynamic parameters obtained from the conformational transition curve are valid, i.e., the equilibrium method which is a noncalorimetric method and DSC measurements yield an identical value of a thermodynamic parameter. The rst two conditions are satised for both proteins, for (i) the CD spectrum of the native protein (unheated solution at 25C) is identical to that of the protein which after denaturation was cooled to 25C, and (ii) the DSC measurements on these proteins provide strong evidence for a two-state transition (7). The third condition can be shown rigorously true if the native and denatured molecules in a two-state transition follow Beer-Lamberts law (8). We have observed that the CD of a protein in the native and denatured states follow Beer-Lamberts law in the concentration range (0.3 0.5 mg/ml) used in this study (results not shown). As far as the fourth condition is concerned, the conventional method of analysis assumes that the pre- and posttransition baselines of the optical transition curve vary linearly with temperature. This will be referred to as the linear model for the description of pre- and posttransition baselines (2). It should be noted (i) that the temperature dependence of y N of both proteins is independent of pH, and (ii) that the temperature dependence of y D of RNase-A showed dependence on pH, and that of lysozyme is independent of pH (see Fig. 1).

210

YADAV AND AHMAD TABLE 1

Assuming a linear dependence of pre- and posttransition baselines on temperature (the linear model), the entire ( y , T ) data of each transition curve of a protein shown in Fig. 1 were tted to the following relation derived from the vant Hoff equation using Eq. [2],
van / R 1/ T 1/ T m y N T y D T exp H m yT , van 1 exp H m / R 1/ T 1/ T m

Thermodynamic Parameters of Lysozyme Using Linear, Mixed-Linear/Parabolic, and Parabolic Models a,b
pH T m, K
van H m , kcal mol 1 cal H m , kcal mol 1

Linear model 1.52 2.00 2.53 3.18 321.3 0.2 [321.1 0.2] 326.1 0.2 [326.2 0.1] 331.6 0.1 [331.4 0.2] 342.4 0.2 [342.2 0.1] 64 3 (70 3) [66 6] 74 3 (77 3) [72 4] 79 2 (84 2) [78 5] 96 3 (100 3) [97 6] 89 96 104 121

[3]
van where H m represents the vant Hoff enthalpy change of denaturation at T m, the temperature at which K D is equal to one. For the linear model y N( T ) and y D( T ) in Eq. [3] are given by relations, y N( T ) a N b N( T ) and y D( T ) a D b D( T ), where a and b are temperatureindependent constants and subscripts N and D represent the native and denatured states, respectively. Thus ( y , T ) data of each transition curve are being t to Eq. [3] with six free parameters, namely H m, T m, van a N, b N, a D, and b D. Values of H m and T m at different pH values for lysozyme and RNase-A, thus obtained, are given in Tables 1 and 2, respectively. These tables cal also contain values of H m , the value of calorimetric van H m. It is seen in Tables 1 and 2 that H m values are determined with an accuracy which is the same as that observed in DSC measurements (210%) (9). It is interesting to note that, as expected for a two-state devan naturation, H m values are in excellent agreement with those obtained from the analysis of denaturation curves followed by observing changes in the absorption properties of these proteins whose pre- and posttransition baselines are described by the linear model (2). A comparison of vant Hoff and calorimetric values of van H m of both proteins suggests that H m values are cal signicantly less than the corresponding H m values (see Tables 1 and 2). It is noteworthy that the value of van H m of tryptophan synthase -subunit from CD denaturation curve is systematically underestimated by cal 15 kcal mol 1 as compared with H m (10). These observations (2; this study) led us to conclude that analysis of a conformational transition curve of a protein assuming a linear temperature dependencies of optical (CD and absorption) properties of N and D molecules van underestimates H m . van A lower value of H m is expected if the heat-induced denaturation of the protein is not a two-state process (11). This possibility is ruled out in our case, for the DSC measurements on these proteins have provided strong evidence for a two-state process (7). A source of discrepancy between vant Hoff and calorimetric H m may stem from the assumption used in the derivation of Eq. [3], namely, that the effects of C p on enthalpy of denaturation are negligible over the entire temperature range of the transition (12). Swint and

Mixed-linear/parabolic model 1.52 2.00 2.53 3.18 322.1 0.2 326.7 0.2 332.2 0.1 342.6 0.1 72 3 (76 3) 83 4 (83 3) 94 2 (92 1) 104 3 (106 3) Parabolic model 1.52 2.00 2.53 3.18 322.0 0.2 326.8 0.2 332.1 0.1 342.5 0.1 84 4 (90 4) 96 5 (97 4) 108 3 (104 2) 116 4 (118 2) 90 97 105 121 90 97 105 121

van a Values of T m and H m are averages of values from three or more van independent measurements at each pH. Values of T m and H m in square brackets were obtained using a different procedure (13), van whereas values of H m in parentheses were taken from Ref. (2). A gives the deviation from the mean of at least three or more independent measurements. cal b H m at a given T m was estimated with the help of the relation, cal 1 H m H ) and C p (kcal mol 1 D C p T m with H D (kcal mol 1 K ) values of 399 and 1.52, which were derived from the data of cal lysozyme given in Ref. (22). The standard errors in H m and C p are 5 and 0.20, respectively (22).

Robertson (12) t the same set of data ( y , T ) of ovomucoid third domain to another relation that assumes the temperature dependence of the enthalpy change on temperature [see Eq. [3] in (12)] and observed that this assumption has no signicant effect on the estimates of van H m of the protein. In order to see whether the t of ( y , T ) data of a transition curve to Eq. [3] is accurate, we have used a different method (13) for the determivan nation of H m and T m. This involves determining (i) the linear temperature dependencies of the pre- and posttransition baselines using the least-squares method; (ii) G D ( RT ln K D) in the range 1.3 G D, kcal mol 1 1.3 using data in the transition region with the help of Eq. [2]; (iii) S m ( ( G D/ T ) p , the entropy change at T m from the linear t of the stability curve (9) (where p is pressure; if the plot of G D versus T is nonlinear over the range of temperature at which protein denatures, S m is then determined from the limited ( G D, T ) data on both sides of T m that fall on a van straight line); and (iv) H m from the product of S m

THERMAL STABILITY OF PROTEINS TABLE 2

211

Thermodynamic Parameters of RNase-A Using Linear, Mixed-Linear/Parabolic, and Parabolic Models a,b
pH T m, K
van H m , kcal mol 1 cal H m , kcal mol 1

Linear model 2.45 3.10 4.48 5.95 313.8 0.2 [313.5 0.1] 322.7 0.1 [323.0 0.2] 334.5 0.1 [334.6 0.2] 336.0 0.1 [336.1 0.2] 71 1 (65 3) [73 5] 77 2 (76 2) [79 5] 89 1 (90 2) [87 5] 92 2 (93 2) [94 4] 86 97 111 113

Mixed-linear/parabolic model 2.45 3.10 4.48 5.95 313.5 0.2 323.3 0.1 334.9 0.1 335.8 0.1 80 4 (77 2) 87 3 (89 3) 98 2 (103 2) 103 2 (106 2) Parabolic model 2.45 3.10 4.48 5.95
a b

86 98 112 113

313.9 0.2 323.3 0.1 335.1 0.1 336.5 0.1

87 3 (83 3) 95 3 (96 3) 106 2 (110 1) 110 2 (113 3)

86 98 112 114

Same as in Table 1. cal H m at a given T m was estimated with the help of the relation, cal 1 H m H ) and C p (kcal mol 1 D C p T m with H D (kcal mol 1 K ) values of 298 and 1.23, which were derived from the data of cal RNase-A given in Ref. (22). The standard errors in H m and C p are 5 and 0.12, respectively (22).

van and T m. Values of H m and T m thus obtained are given in square brackets in Tables 1 and 2, where it is seen that these values are, within experimental errors, identical to those obtained from the t of entire transition data to Eq. [3] using the linear model. This van agreement between estimates of H m and T m obtained by two different methods (13, 14) led us to believe that the analysis of the transition data according to Eq. [3] is accurate, and the source of discrepancy van cal between H m and H m of a protein at any pH is not due to the assumption that the effects of C p on enthalpy change are negligible over the transition zone. Tiktopulo and Privalov (15) have studied the thermal-induced denaturation of RNase-A at different pH values using optical techniques and DSC measurements. They have shown that the most symmetrical sigmoidal normalized conformational transition curve, a characteristic of a two-state process, is obtained only when a linear function for the posttransition baseline ( y D( T ) a D b DT ) and a parabolic function (linear for the rst derivative) for the pretransition baseline ( y N( T ) a N b NT c NT 2 ) are used for the extrapolations of post- and pretransition baselines, respec-

tively, into the transition region. A comparison of van cal H m thus obtained with H m suggested that the agreement between these estimates is within 10% for RNase-A (15). We have, therefore, analyzed each transition curve shown in Fig. 1, according to Eq. [3] using the mixed-linear/parabolic model, i.e., the temperature dependencies of y D( T ) and y N( T ) described by the linear and second degree polynomial functions, respectively. This analysis which involves tting seven free parameters ( H m, T m, a N, b N, c N, a D, and b D) to Eq. [3] van gave values of H m and T m that are given in Tables 1 and 2 for lysozyme and RNase-A, respectively. It is seen in these tables that although the agreement bevan cal tween H m and H m of a protein at a given pH is van cal better than H m and H m obtained from the analysis cal of the same set of data using linear model, H m of a protein at each pH is still signicantly higher. For a protein the disagreement between the vant Hoff and calorimetric H m obtained from the analysis of the same set of data according to Eq. [3] using the linear model as well as the mixed-linear/parabolic model led us to investigate the effect of temperature on [] 222 of the denatured proteins. The results of these measurements of a protein are shown in the inset of Fig. 1 where it is seen that the CD property of the oxidized RNase-A at 222 nm is indeed nonlinear in temperature. It has been observed that the change in this optical property of the denatured protein is described adequately by a second degree polynomial equation in temperature. We have therefore analyzed each heat-induced denaturation curve of both proteins according to Eq. [3] with the parabolic model, i.e., the temperature dependencies of both y N( T ) and y D( T ) are described by a polynomial equation. The least-squares analysis which involves tting eight free parameters ( H m, T m, a N, b N, c N, a D, b D, and c D) gave values of H m and T m of a protein at a given pH. These values van are given in Tables 1 and 2. A comparison of H m thus cal obtained with H m of a protein at each pH suggests that the agreement is excellent. This agreement provides strong support for the analysis of thermal transition curves followed by measuring changes in the chiroptical property ([] 222), using the parabolic model. It is interesting to note that the same conclusion has been reached from the analysis of heat-induced denaturation measured by the absorption properties of proteins using the parabolic model to describe the dependencies of y N( T ) and y D( T ) on temperature (2). In this paper, a smooth denaturation curve is being t to Eq. [3] with several parameters which are oated simultaneously. A question that arises is whether such van a t gives unique values of H m and T m. We have checked the accuracy of these thermodynamic parameters in the case of the linear baseline model using a different procedure (13) which analyses ( y , T ) data in the pre- and posttransition regions and those in the

212

YADAV AND AHMAD

van FIG. 2. Plots of H m of lysozyme (A) and RNase-A (B) versus corresponding T m at different pH values. For each protein, the symbols used in this gure related the pH values that are shown in Fig. 1. Values of thermodynamic parameters were obtained according to Eq. [3] with the linear (1), mixed-linear/parabolic (2), and parabolic (3) models used to describe the temperature dependencies of y N and y D.

transition region in two steps (see above). It has been van observed that values of H m and T m obtained by the latter method (13) are in excellent agreement with those by tting the entire transition curve to Eq. [3] with six free parameters namely, H m, T m, a N, b N, a D, and b D. This agreement between the estimates of a thermodynamic parameters from two different procedures led us to believe that our method of analysis of denaturation curve is accurate. It should, however, be noted that the method involving analysis in two steps (13) cannot be used in the cases of mixed-linear/parabolic and parabolic baseline models. The reason for saying this is that this procedure (13) involves the division of a smooth denaturation curve into three well-dened regions, namely, pretransition, transition, and posttransition regions, and this division requires some criterion that must be adopted to decide which points represent curvature due to inherent temperature dependence of CD properties rather than onset or completion of the unfolding transition. We, however, believe that analysis which ts the entire ( y , T ) data to Eq. [3] with eight free parameters is accurate not only because of the excellent agreement that exists between equilibrium and calorimetric estimates of a thermodynamic parameter but also because of our observation

that the temperature dependence of baseline data is indeed nonlinear. There are several methods that have been used to determine C p of a protein from the measurements of its conformational transition curve (12, 16 20). However, the recommended procedure involves the measurements of heat-induced denaturation of the protein at different pH values and the estimation of C p from van the linear plot of H m versus temperature (7, 9), for C p of the protein does not depend on pH and tempervan ature in the range 20 80C (7, 12, 20, 21). ( H m , T m) data of Tables 1 and 2 are shown in Fig. 2 as plots of van H m versus corresponding T m of lysozyme and RNase-A, respectively. For each protein, curves 1, 2, van and 3 represent ( H m , T m) data obtained from the t of its heat-induced transition curves (Fig. 1) to Eq. [3] using the linear, mixed-linear/parabolic, and parabolic models, respectively. It is seen in Fig. 2 that, for a protein, a linear least-squares analysis of each set of data gave values of C p which are, within experimental errors, identical to one another, i.e., C p determined from the analysis of a transition curve is independent of the model used to describe the temperature dependencies of the pre- and posttransition baselines. Furthermore, values of C p of both the proteins, ob-

THERMAL STABILITY OF PROTEINS

213

tained from our studies, are in excellent agreement with those obtained by DSC measurements (22); values of calorimetric C p (kcal mol 1 K 1) are 1.52 0.20 for lysozyme and 1.23 0.12 for RNase-A. A comparison of errors involved in the determination of calorimetric C p and that from the analysis of the transition curves (see Fig. 2) suggests that C p can be measured with the same accuracy as that with a calorimeter. It is noteworthy that Sinha et al. (2) have estimated C p from the heat-induced transition curves of lysozyme and RNase-A followed by observing changes in their absorption properties. They have also reached the same conclusions, namely, (i) C p determined from the heat-induced denaturation curve is independent of the model used to describe the pre- and posttransition baselines, (ii) C p values of these proteins are in excellent agreement with those from DSC measurements, and (iii) C p values are measured with the same error as that observed with a calorimeter. o Finally, the value of G D of a protein can be accurately estimated using the Gibbs-Helmoltz equation,
o van G D H m

ACKNOWLEDGMENTS
This research was supported by grants from the Department of Science and Technology and Council of Scientic and Industrial Research (India).

REFERENCES
1. Pfeil, W. (1998) in Protein Stability and Folding: A Collection of Thermodynamic Data, Springer, Berlin. 2. Sinha, A., Yadav, S., Ahmad, R., and Ahmad, F. (2000) Biochem. J. 345, 711717. 3. Ahmad, F. (1991) Ind. J. Biochem. Biophys. 28, 168 173. 4. Hamaguchi, K., and Kurono, A. (1963) J. Biochem. 54, 111122. 5. Bigelow, C. C. (1964) J. Mol. Biol. 8, 696 701. 6. Chen, Y. D., Yang, J. T., and Martinez, H. M. (1972) Biochemistry 11, 4120 4126. 7. Privalov, P. L. (1979) Adv. Protein Chem. 33, 167241. 8. Klotz, I. M. (1967) in Energy Changes in Biochemical Reactions, Academic Press, New York. 9. Becktel, W. J., and Schellman, J. A. (1987) Biopolymers 26, 1859 1877. 10. Ogasahara, K., Yutani, K., Suzuki, M., and Sugino, Y. (1984) Int. J. Peptide Protein Res. 24, 147154. 11. Tanford, C. (1968) Adv. Protein Chem. 23, 121282. 12. Swint, L., and Robertson, A. D. (1993) Protein Sci. 2, 20372049. 13. Taneja, S., and Ahmad, F. (1994) Biochem. J. 303, 147153. 14. Santoro, M. M., and Bolen, D. W. (1992) Biochemistry 31, 4901 4907. 15. Tiktopulo, E. I., and Privalov, P. L. (1974) Biophys. Chem. 1, 349 357. 16. Brandts, J. F., and Hunt, L. (1967) J. Am. Chem. Soc. 89, 4826 4838. 17. Pace, C. N., and Tanford, C. (1968) Biochemistry 7, 198 208. 18. Shiao, D. F., Lumry, R., and Fahay, J. (1971) J. Am. Chem. Soc. 93, 2024 2035. 19. Nojima, H., Ikai, A., Oshima, T., and Noda, H. (1977) J. Mol. Biol. 116, 429 422. 20. Pace, C. N., and Laurent, D. V. (1989) Biochemistry 28, 2520 2525. 21. Griko, Y. V., and Privalov, P. L. (1992) Biochemistry 31, 8810 8815. 22. Privalov, P. L., and Gill, S. J. (1988) Adv. Protein Chem. 39, 191234.

T m 298.15 C p T m 298.15 Tm 298.15 ln

298.15 Tm

[4]

van if values of H m , T m, and C p are accurately known from the analysis of the optical transition curves. Here we propose the following method for the determination o of G D , the protein stability from a two-state heatinduced denaturation curve of a protein. (i) Measure thermal denaturation curves followed by observing changes in [] 222 at different pH values, (ii) determine van H m and T m values by tting each denaturation curve to Eq. [3] using the parabolic model, i.e., with eight free parameters ( H m, T m, a N, b N, c N, a D, b D, and c D), (iii) van estimate C p from the plot of H m versus T m, and (iv) o determine G D using Eq. [4].